PowerOP, Whatchanged/Issues/TODO 2/2

PowerOP, Whatchanged/Issues/TODO 2/2

[Eugeny S. Mints]

whatchanged:
- optional sysfs interface is added
- code is moved under kernel/power
- Greg's comments on kobject and coding style are addressed

todo/issues:
- better implementation for getting registered operating point names
- move string parsing into powerop generic code
- configfs for operating points creation from user space

Re: PowerOP, Whatchanged/Issues/TODO 2/2

[Greg KH]

On Sat, Sep 02, 2006 at 07:16:01PM +0400, Eugeny S. Mints wrote:
> whatchanged:
> - optional sysfs interface is added
> - code is moved under kernel/power
> - Greg's comments on kobject and coding style are addressed
> 
> todo/issues:
> - better implementation for getting registered operating point names
> - move string parsing into powerop generic code
> - configfs for operating points creation from user space

You forgot:
  - Handle module referencing counting issues.

Also, no #ifdefs in the .c code please, that debugging stuff doesn't
have to look like that, break it out into separate functions to keep the
codepath clean.

thanks,

greg k-h

Re: PowerOP, Whatchanged/Issues/TODO 2/2

[Eugeny S. Mints]

Greg,

Greg KH wrote:
> On Sat, Sep 02, 2006 at 07:16:01PM +0400, Eugeny S. Mints wrote:
>> whatchanged:
>> - optional sysfs interface is added
>> - code is moved under kernel/power
>> - Greg's comments on kobject and coding style are addressed
>>
>> todo/issues:
>> - better implementation for getting registered operating point names
>> - move string parsing into powerop generic code
>> - configfs for operating points creation from user space
> 
> You forgot:
>   - Handle module referencing counting issues.
i'm discussing this and other issues you pointed out in a separate email.
> 
> Also, no #ifdefs in the .c code please, that debugging stuff doesn't
> have to look like that,
do you mean CONFIG_POWEROP_SYSFS_OP_DEBUG_IF? Probably it's not a good name but 
it's not a debugging code. it could be CONFIG_POWEROP_SYSFS_OP_HW_IF - this 
special HW point might be useful in a production code.

Eugeny
> break it out into separate functions to keep the
> codepath clean.
> 
> thanks,
> 
> greg k-h
> 

Re: PowerOP, Whatchanged/Issues/TODO 2/2

[Greg KH]

On Sun, Sep 03, 2006 at 03:13:03AM +0400, Eugeny S. Mints wrote:
> Greg,
> 
> Greg KH wrote:
> >On Sat, Sep 02, 2006 at 07:16:01PM +0400, Eugeny S. Mints wrote:
> >>whatchanged:
> >>- optional sysfs interface is added
> >>- code is moved under kernel/power
> >>- Greg's comments on kobject and coding style are addressed
> >>
> >>todo/issues:
> >>- better implementation for getting registered operating point names
> >>- move string parsing into powerop generic code
> >>- configfs for operating points creation from user space
> >
> >You forgot:
> >  - Handle module referencing counting issues.
> i'm discussing this and other issues you pointed out in a separate email.
> >
> >Also, no #ifdefs in the .c code please, that debugging stuff doesn't
> >have to look like that,
> do you mean CONFIG_POWEROP_SYSFS_OP_DEBUG_IF? Probably it's not a good name 
> but it's not a debugging code. it could be CONFIG_POWEROP_SYSFS_OP_HW_IF - 
> this special HW point might be useful in a production code.

No matter what the name, the #ifdef does not belong in the .c code in
the manner like this was done.  Please move it.

thanks,

greg k-h

Re: PowerOP, Whatchanged/Issues/TODO 2/2

[Pavel Machek]

Hi!

> whatchanged:
> - optional sysfs interface is added
> - code is moved under kernel/power
> - Greg's comments on kobject and coding style are addressed
> 
> todo/issues:
> - better implementation for getting registered operating point names
> - move string parsing into powerop generic code

Can we get rid of string parsing altogether? It should be possible to
design better in-kernel interface than char *. Can you just use opaque
structure?

-- 
Thanks for all the (sleeping) penguins.

PowerOP, Whatchanged/Issues/TODO 2/2

[Eugeny S. Mints]

whatchanged:
- optional sysfs interface is added
- code is moved under kernel/power
- Greg's comments on kobject and coding style are addressed

todo/issues:
- better implementation for getting registered operating point names
- move string parsing into powerop generic code
- configfs for operating points creation from user space
- powerop driver module refcounting

PowerOP, Whatchanged/Issues/TODO 2/2

[Eugeny S. Mints]

whatchanged:
- new concept of powerop power parameter is introduced
- new powerop core interface routine to set individual(or subset of)
  power parameter value is added
- no more string parsing at any layer
- no more va_list interface 
- powerop driver module refcounting is added
- documentation file including optional sysfs interface description is added

todo/issues:
- better implementation for getting registered operating point names
- configfs for operating points creation from user space